Principal Investigator

Cigarette smoking remains a major health problem in the U.S. and worldwide. The observed overall decline in smoking prevalence in the U.S. during 2005-2010 is not mirrored by a similar decline in teen smoking. The goal of this project is to develop and implement a theoretical network formation model that will enable a better characterization of the environment in which teens make decisions concerning risky activities. In particular, this project initiates a comprehensive study on how friendship networks affect individual choices and, conversely, how individual decisions to engage in risky activities affect friendship networks. The proposed approach differs markedly from previous work in that it models simultaneously decisions regarding risky activities and decisions regarding friend selection, within a single framework. In the proposed model for studying peer effects, causality can flow in both directions: individuals’ choices affect selection of friends and friends’ behavior (the proportion of current friends who smoke) affects individuals’ choices (to smoke or not). This research also develops an innovative way to handle multiplicity of equilibria (multiple possible network and action configurations) – a ubiquitous problem in research in social sciences. Specifically, the behavioral assumptions of the model we develop naturally assign probabilities to all equilibria of the static game (in contrast to previous work which adopts somewhat ad hoc selection rules). Finally, the model has the capability of generating cascade effects – a network feature where a large proportion of the population finds it optimal to alter their behavior (become smokers or engage in a risky activity for example) over a short timeframe. Previous models in the literature consider only network formation or teen choices given an exogenous network and cannot generate cascade effects. The empirical work will examine the evidence for cascade effects in the context of smoking behavior and evaluate the epidemic hazard, (i.e. the probability of a cascade effect happening) which depends on model parameters. The proposed TRIO grant will enable further development of the broader research agenda on understanding the determinants of socioeconomic networks and individual choices with prospects of attracting future external funding. It will build on substantial progress that has already been made in developing the theoretical model, building an estimation strategy, and obtaining approval for the restricted-use Add-Health dataset.

Funded By
Award Dates